OpenAI has revealed its first custom-designed processor for artificial intelligence, developed in partnership with Broadcom. Named Jalapeño, the chip is an application-specific integrated circuit (ASIC) built specifically for AI inference — the process by which models respond to user requests in products like ChatGPT or Codex.
The announcement comes nine months after OpenAI first disclosed its collaboration with Broadcom on chip development. The rapid timeline from design to manufacturing tape-out was reportedly aided by OpenAI's own AI models, marking a significant step in the company's effort to reduce reliance on off-the-shelf processors from rivals like Nvidia.
Jalapeño is an inference-focused accelerator, meaning it optimizes the workload of running trained models rather than training new ones. Early testing suggests the first-generation chip will deliver performance per watt substantially better than current state-of-the-art alternatives, according to OpenAI.
By controlling its own silicon, OpenAI aims to lower costs and improve efficiency for its growing suite of AI services. The move also positions the company to better tailor hardware to the evolving demands of its large language models, potentially giving it a competitive edge in the AI infrastructure race.
Broadcom's role as manufacturing partner underscores the growing collaboration between AI firms and established chip designers. The Jalapeño name continues OpenAI's tradition of spicy codenames for internal projects.